Proforma / Research & field notes
Understand the work.
Account for the value.
Research on R&D tax, engineering evidence and the economics of AI. Written for the people building the work—and the people reviewing it.
October 2026 edition
15 research articles
Primary sources. Practical methods.
Field guide
Tax changes
What changed, when it applies and which decisions need attention.
Australia’s R&D Tax Incentive in 2026: current rules, proposed reforms and the records to prepare
Separate Australia’s current R&DTI rules from the proposed 2028 reforms, then build a practical activity, cost and evidence review.
Read articleFederal R&D relief does not finish your state tax work
California shows why federal Section 174A balances, state deductions and research credits need separate evidence and calculations.
Read articleForm 6765 in 2026: build the project-to-cost map before filing season
A practical guide to Section G's effective tax years, reporting exceptions and the evidence map engineering and finance should prepare.
Read articleOne engineering team, two tax systems: organizing US–Australia R&D costs
A practical evidence model for engineering teams spanning the United States and Australia, with separate entity, location, contract and period decisions.
Read articleSection 174A changed R&D expensing. Your evidence still has to connect.
What the 2025 domestic R&D deduction change means for engineering cost records, transition balances and cash-flow analysis in 2026.
Read articleThe $500,000 R&D payroll credit headline is a limit, not a cash forecast
How startup research credits move from qualified work to an election, payroll-tax use and a realistic runway model.
Read articleField guide
AI & ROI
Measure engineering value with costs, quality and a credible comparison.
AI coding ROI: measure the work that survives review
A practical way to measure AI coding returns using completed outcomes, review effort and real costs, informed by studies that found very different effects.
Read articleAI evaluation economics: measure cost per accepted outcome
Compare AI workflows using accepted outcomes, retries, human review and failure costs, with an explicit experiment budget and quality threshold.
Read articleConnect AI and cloud spend to the engineering work it paid for
Build a reviewable trail from an AI or cloud invoice to usage, experiments and project decisions, without confusing allocation with tax eligibility.
Read articleField guide
Engineering evidence
Connect technical questions, experiments, work records and spend.
Does your AI project contain R&D? An Australian evidence walkthrough
Use a worked AI engineering example to separate experimentation, supporting work and routine delivery under Australia’s current R&DTI guidance.
Read articleGitHub is a starting point for R&D evidence, not the whole story
A practical evidence chain from technical questions to experiments, decisions and costs, including GitHub’s October 2026 retention change.
Read articleUse AI to draft the R&D narrative. Make every important claim inspectable.
A practical method for AI-assisted technical narratives that preserves source versions, separates inference from fact, and makes missing evidence visible.
Read articleWhen does AI software work become qualified research? Start with the uncertainty.
An evidence-first way to distinguish software experiments from routine integration, with a worked AI retrieval example and questions for a tax reviewer.
Read articleField guide
Adviser practice
Prepare a review that another person can follow and challenge.
The adviser’s R&D evidence intake: fewer follow-ups, clearer review and measurable ROI
A practical source-to-workpaper intake method, with an illustrative cost model for reviewer time, client effort and rework.
Read articleWho can claim outsourced R&D? Follow the contract, rights and research risk.
Why a development invoice is not automatically a research-credit expense, and how customers and contractors can prepare the rights, risk and activity evidence.
Read articleInside the platform
Notes from building Proforma.
Architecture, working methods and decisions from the product journal.